Hi, Andrii
On 2022/7/6 13:05, Andrii Nakryiko wrote:
> On Mon, Jul 4, 2022 at 7:09 AM Hengqi Chen <hengqi.chen@xxxxxxxxx> wrote:
>>
>> The binary_path parameter is required for bpf_program__attach_usdt().
>> Error out when user attach USDT probe without specifying a binary_path.
>>
>
> This is a required parameter, libbpf doesn't add pr_warn() for every
> `const char *` parameter that the user incorrectly passes NULL for
> (e.g., bpf_program__attach_kprobe's func_name). If you think
I understand this is a required parameter. The intention of this patch is
to avoid coredump if user passes NULL for binary_path argument, not just
emit a warning. The uprobe handling code of libbpf already did this.
BTW, most of libbpf APIs do NULL check for their const char * parameters
and return -EINVAL.
> bpf_program__attach_usdt() doc comment about this is not clear enough,
> let's improve the documentation instead of littering libbpf source
> code with Java-like NULL checks everywhere.
>
>> Signed-off-by: Hengqi Chen <hengqi.chen@xxxxxxxxx>
>> ---
>> tools/lib/bpf/libbpf.c | 6 ++++++
>> 1 file changed, 6 insertions(+)
>>
>> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
>> index 8a45a84eb9b2..5e4153c5b0a6 100644
>> --- a/tools/lib/bpf/libbpf.c
>> +++ b/tools/lib/bpf/libbpf.c
>> @@ -10686,6 +10686,12 @@ struct bpf_link *bpf_program__attach_usdt(const struct bpf_program *prog,
>> return libbpf_err_ptr(-EINVAL);
>> }
>>
>> + if (!binary_path) {
>> + pr_warn("prog '%s': USDT attach requires binary_path\n",
>> + prog->name);
>> + return libbpf_err_ptr(-EINVAL);
>> + }
>> +
>> if (!strchr(binary_path, '/')) {
>> err = resolve_full_path(binary_path, resolved_path, sizeof(resolved_path));
>> if (err) {
>> --
>> 2.30.2
--
Hengqi
